Site selection method for compressed hydrogen energy storage underground reservoir

Abstract

The invention relates to the technical field of underground reservoir site selection, in particular to a compressed hydrogen energy storage underground reservoir site selection method, which comprises the following steps: collecting multi-source drilling data of each candidate reservoir site, and extracting microscopic lithologic characteristics, macroscopic stratum characteristics and parameter correlation characteristics; constructing a geological architecture model according to the macroscopic stratum characteristics, and generating a plurality of heterogeneous random geological models by taking data reflecting the microcosmic lithologic characteristics and the parameter correlation characteristics as soft data; selecting a plurality of models as representative models, and performing multi-physics field coupling numerical simulation to obtain a series of performance index sets; performing uncertainty quantification processing on each performance index set, and performing comprehensive score sorting on each candidate library address by adopting a multi-objective optimization sorting method; according to the method, multi-source drilling data can be fully utilized, geological uncertainty is considered, high-fidelity random geological modeling and simulation can be carried out, and quantitative evaluation can be carried out on the performance and the risk of a plurality of candidate library addresses.

Description

Technical Field

[0001] This invention relates to the field of underground storage site selection technology, and specifically to a method for selecting a site for an underground compressed hydrogen energy storage facility. Background Technology

[0002] Compressed hydrogen energy storage is a key technology for realizing large-scale, long-term hydrogen energy storage. It places extremely stringent requirements on the sealing, stability, and injection and extraction efficiency of underground storage facilities. An ideal underground storage facility for compressed hydrogen energy storage must have extremely low permeability to prevent hydrogen leakage, good mechanical stability to withstand cyclic injection and extraction pressure, sufficient pore volume or solubility to provide effective storage capacity, and favorable geological structure to avoid communication with faults or aquifers. Therefore, the site selection of underground storage facilities for compressed hydrogen energy storage is particularly critical.

[0003] The site selection for underground compressed hydrogen energy storage typically involves borehole data analysis, geological modeling, and site evaluation. Existing site selection methods, on the one hand, rely on relatively simplistic use of borehole data. First-hand data such as borehole cores and logging curves are usually only manually layered and simply interpreted, failing to deeply explore the implicit micro- and macro-level characteristics closely related to reservoir performance, such as lithological sequences, spatial variations in mineral composition, statistical regularities of microfracture development, and indicative information on geostress direction. On the other hand, geological modeling is often idealized, with the constructed geomechanical models mostly being homogeneous layered models or based on simple interpolation. The current model ignores the randomness and structural variation of rock mechanics and seepage parameters in three-dimensional space. This smooth model cannot characterize local weak zones, high-permeability strips or stress anomaly zones in actual strata, leading to overly optimistic or blind-zone-based simulation results of stability and sealing performance. In addition, the existing site selection methods are relatively one-sided in their site evaluation, often using a single index or a simple weighted scoring method. The weights of each index are subjectively set, and there is a lack of dynamic and quantitative simulation verification that combines geological models with real multiphysics processes. This makes it impossible to predict the performance evolution and potential risks of reservoirs under long-term cyclic operation.

[0004] Therefore, it is necessary to provide a method for selecting underground storage sites for compressed hydrogen energy storage, which can overcome the shortcomings of existing underground storage site selection methods that rely on experience, have distorted geological modeling, and have one-sided site selection evaluation. This method can make full use of borehole data, perform high-fidelity geological modeling and simulation, and provide quantitative and comparative evaluation of the sealing, injectability, and long-term stability of multiple candidate sites, thus scientifically guiding the optimal site selection for underground storage. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides a method for selecting a site for an underground compressed hydrogen energy storage facility and its application, thereby resolving the technical issues raised in the background section.

[0006] The technical solution adopted by this invention to solve its technical problem is as follows: A method for site selection of an underground compressed hydrogen energy storage facility includes the following steps: S1: Collect multi-source borehole data for each candidate reservoir site, including high-resolution core scanning images, digital logging curve sequences, and core laboratory test data, and extract the microscopic lithological characteristics, macroscopic stratigraphic characteristics, and parameter correlation characteristics of each candidate reservoir site from them. S2: For each candidate reservoir site, a corresponding geological framework model is constructed based on its macroscopic stratigraphic characteristics. Within the corresponding geological framework model, the sequential Gaussian co-simulation method is used, with its core laboratory test data as hard data and its data reflecting microscopic lithological characteristics and parameter correlation characteristics as soft data, to generate several heterogeneous stochastic geological models containing stochastic parameter fields. S3: For each candidate site, select multiple representative models from its heterogeneous random geological models, perform multiphysics coupled numerical simulations on the representative models, calculate the performance indices of different representative models, and form a performance index set for each identical performance index of different representative models. S4: For each candidate repository, perform uncertainty quantification on its performance index set. Based on the uncertainty quantification result, use a multi-objective optimization ranking method to comprehensively score and rank each candidate repository to obtain the preferred repository with the highest ranking.

[0007] Furthermore, the specific steps of S3 are as follows: S31: The geological architecture model is divided into grids, and the random parameter field of each representative model is mapped to each grid cell of the geological architecture model to obtain the corresponding simulation parameter field model. S32: Establish the solid deformation relationship, gas permeation relationship, and component diffusion relationship for each representative model to obtain the corresponding set of governing equations; S33: Based on the simulation parameter field model, the sequential coupling iterative method is used to solve the corresponding control equations of each representative model within the complete simulation cycle to obtain the corresponding coupled simulation results. S34: Post-process the corresponding coupled simulation results of each representative model to obtain different performance indices. Each identical performance index of different representative models forms a performance index set.

[0008] Furthermore, the method for establishing the solid deformation relationship in S32 is specifically as follows: The equilibrium equations are established using the incremental small-deformation assumption method, and the equilibrium equations are as follows: In the formula, The effective stress tensor is given by p, the pore pressure is given by α, the Biot coefficient is given by I, and the unit tensor is given by I. Let g be the volume density, and g be the gravitational acceleration vector. Establish the displacement-strain geometric equations: In the formula, ε is the strain tensor and u is the displacement vector; Based on the elastoplastic constitutive relation, the stress-strain relationship between the strain tensor and the effective stress tensor is established, and the solid deformation relationship between the associated displacement and the effective stress is obtained.

[0009] Furthermore, the method for establishing the gas permeation relationship in S32 is specifically as follows: Establish the percolation-compressibility correlation equation for hydrogen gas in porous media: In the formula, φ represents porosity. Let be the density of hydrogen gas, t be the time variable, and v be Darcy's velocity vector, satisfying Darcy's law formula. μ is the dynamic viscosity of hydrogen gas, and K is the permeability. This represents the source and sink terms of the injection and production wells; Substituting Darcy's law into the seepage-compressibility correlation equation, we obtain a nonlinear gas seepage relationship with pore pressure p as the variable: Among them, the change in porosity φ is related to volumetric strain. It is related to the pore pressure p, that is, , It is the bulk modulus of solid particles.

[0010] Furthermore, the method for determining the component diffusion relationship in S32 specifically involves establishing the component diffusion relationship based on Fick's law governing the molecular diffusion of hydrogen in residual formation water: ; In the formula, Let C be the water saturation, C be the hydrogen concentration in the water, and D be the effective diffusion coefficient tensor. R represents the aqueous phase seepage velocity, and R is the chemical reaction source term, with a value of 0.

[0011] Furthermore, the sequential coupling iterative method in S33 specifically includes: Set boundary conditions and operating conditions; A complete simulation cycle is divided into several time steps, and multiple iterations are performed in each time step. Each iteration fixes the displacement field obtained in the previous iteration at the current time step, and solves the problem by coupling the gas permeation relationship and the component diffusion relationship, and iteratively updates the pore pressure field and hydrogen concentration field at the current time step. Using the updated pore pressure field and hydrogen concentration field, the solid deformation relationship is solved, and the displacement field at the current time step is updated. Iterate multiple times at the current time step until convergence, then proceed to the next time step; The coupled simulation results obtained after iteratively solving the output control equations include the pore pressure field, displacement field, and hydrogen concentration field that vary with the time step.

[0012] Furthermore, the specific method for post-processing calculation in S34 is as follows: The velocity vector field is calculated based on the pore pressure field, and the sealing index is calculated by combining the total hydrogen injection mass. During the simulation solution process, the wellhead pressure-time curve and hydrogen flow-time curve are output. The average hydrogen injection rate is extracted from them, and the average pressure difference between the wellhead and the reservoir is extracted from the digital logging curve sequence. The injectability index is then calculated. The maximum equivalent plastic strain and the rate of change of cavity volume are calculated based on the displacement field throughout the entire simulation period. Based on the pore pressure field, displacement field, and hydrogen concentration field, the component flow rate at the extraction end is tracked during the simulation process, and the hydrogen purity maintenance index is calculated.

[0013] Furthermore, the specific method for uncertainty quantification in S4 is as follows: the mean, standard deviation and probability distribution of each performance index set are statistically analyzed to obtain the uncertainty quantification results of different performance indices.

[0014] Furthermore, the multi-objective optimization ranking method in S4 specifically involves minimizing the sealing performance index, minimizing the injectability index, minimizing the stability index, and maximizing the hydrogen purity maintenance index as optimization objectives. Using a ranking method that approximates the ideal solution, and combining the decision-maker's preferences, a comprehensive score is given for multiple performance indicators of each candidate site. A risk assessment is then performed on the top-ranked preferred site, providing a prediction of its average performance indicators and a risk probability distribution map for each performance indicator.

[0015] Furthermore, in S1, The specific method for extracting microscopic lithological features is as follows: a convolutional neural network is used to train high-resolution scanned images of rock cores to automatically identify and extract microscopic lithological features; The specific method for extracting macroscopic stratigraphic features is as follows: Long Short-Term Memory (LSTM) network is used to analyze continuous digital logging curve sequences to automatically identify macroscopic stratigraphic features; The specific method for extracting parameter correlation features is as follows: principal component analysis is performed on the core laboratory test data and the corresponding logging response data in the digital logging curve sequence to extract the comprehensive characteristic factors that can best represent the changes in rock physical and mechanical properties as parameter correlation features.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention provides a method for site selection of underground compressed hydrogen energy storage facilities. It combines random field geological modeling with multiphysics field coupled simulation, fully considering the randomness of geological modeling in three-dimensional space. This generates a heterogeneous random geological model with strong heterogeneity, clear geological significance, and greater realism. Multiphysics field coupled simulation is performed on multiple heterogeneous random geological models to simulate the real physical process of storage facility operation. A series of performance indexes are obtained. Each performance index is not unique, but rather has corresponding performance indices for different heterogeneous random geological models. The simulation results are more comprehensive and realistic, and the performance indices are all physical quantities that can be directly used for engineering design, providing greater guidance for site selection assessment.

[0017] 2. The present invention provides a method for selecting underground storage sites for compressed hydrogen energy storage. It utilizes convolutional neural networks, long short-term memory networks, and principal component analysis to extract microscopic lithological features, macroscopic stratigraphic features, and correlation parameter features from multi-source borehole data, respectively. This method deeply mines and quantifies multi-source borehole data, greatly improving the accuracy of geological understanding of candidate storage site areas.

[0018] 3. The present invention provides a method for site selection of underground compressed hydrogen energy storage, which quantifies the uncertainty of the performance index set obtained by random simulation, and directly transmits the geological uncertainty to the engineering performance prediction. The decision result is no longer a single value, but a probability distribution, making the risk of the storage site area predictable, the site selection decision more scientific, transparent and traceable, and greatly reducing human intervention and subjective bias. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the site selection method for an underground compressed hydrogen energy storage facility as described in this embodiment; Figure 2 for Figure 1 A flowchart of the S3 process; Figure 3 A schematic diagram of the macroscopic stratigraphic features of two candidate reservoir sites, A and B. Figure 4 This is a schematic diagram of the simulation results of a representative model corresponding to two candidate library site regions, A and B. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] This invention provides a method for site selection of underground compressed hydrogen energy storage facilities, such as... Figure 1 As shown, the method includes the following steps: S1: Collect multi-source borehole data for each candidate reservoir site, including high-resolution core scanning images, digital logging curve sequences, and core laboratory test data, and extract the microscopic lithological characteristics, macroscopic stratigraphic characteristics, and parameter correlation characteristics of each candidate reservoir site from them. Furthermore, the method for extracting the microscopic lithological features is as follows: a convolutional neural network is used to train the high-resolution scanned image of the rock core to automatically identify and quantify the density, direction, aperture, mineral grain distribution, and bedding plane development of the rock core as microscopic lithological features. The microscopic lithological features reflect the microscopic lithological structure and composition characteristics of the rock core. The method for extracting macroscopic stratigraphic features is as follows: Long Short-Term Memory Network or one-dimensional convolutional neural network is used to analyze the continuous digital logging curve sequence, automatically identify key lithological interfaces and sedimentary cycles in the candidate reservoir area, and detect potential weak interlayers or fracture zones with abnormal logging responses, so as to obtain macroscopic stratigraphic features that reflect the macroscopic stratigraphic sequence and interface characteristics of the candidate reservoir area. In this embodiment, the extraction method is preferably Long Short-Term Memory Network. The specific method for extracting the parameter correlation features is as follows: principal component analysis or autoencoder processing is performed on the corresponding logging response data in the core laboratory test data and digital logging curve sequence to extract several comprehensive feature factors that can best represent the changes in the physical and mechanical properties of the rock as parameter correlation features. For example, the acoustic-density comprehensive feature factor is extracted as a parameter correlation feature to reflect the compactness and porosity of the rock, and the resistivity-natural gamma comprehensive feature factor is extracted as a parameter correlation feature to reflect the clay content and fluid properties. The parameter correlation features remove redundant data and reduce the feature dimension of the original multi-source borehole data while retaining as much information as possible, which is convenient for subsequent random field modeling. In this embodiment, the extraction method is preferably principal component analysis. By using different artificial intelligence or machine learning algorithms, microscopic lithological features, macroscopic stratigraphic features, and correlation parameter features can be automatically extracted from multi-source borehole data. This enables in-depth mining and quantification of multi-source borehole data, greatly improving the accuracy of geological understanding of candidate reservoir sites.

[0023] S2: For each candidate reservoir site, a corresponding geological framework model is constructed based on its macroscopic stratigraphic characteristics. Within the corresponding geological framework model, the sequential Gaussian co-simulation method is used, with its core laboratory test data as hard data and its data reflecting microscopic lithological characteristics and parameter correlation characteristics as soft data, to generate several heterogeneous stochastic geological models containing stochastic parameter fields. Furthermore, by utilizing key lithological interfaces in macroscopic stratigraphic features and combining them with existing regional geological interpretation results, a deterministic geological framework model can be constructed. These regional geological interpretation results are obtained through existing regional geological survey methods, such as regional stratigraphic correlation columnar sections, which indicate the thickness and vertical relationship of each lithological stratum, and regional tectonic outline maps, which indicate the distribution and properties of major faults and folds. The geological framework model is a three-dimensional geometric model that includes reservoirs, caprocks, and floor, which can reflect the stratigraphic framework and major faults of the candidate reservoir site area, providing preliminary constraints for the generation of heterogeneous stochastic geological models. The specific method for generating heterogeneous stochastic geological models is as follows: For different key parameters related to the selection of candidate pool sites, including permeability, elastic modulus, tensile strength, etc., within the determined geological framework model, the sequential Gaussian collaborative simulation method is used sequentially. Known data of the corresponding key parameters in sparse core laboratory test data are used as hard data. For example, for the unknown permeability of a certain area, the known permeability obtained from the core laboratory test data of the neighboring area is used as hard data. Data reflecting the microfracture density characteristics and parameter correlation characteristics in the microlithological characteristics are used as soft data, thereby providing information on the continuous variation trend of different key parameters in the geological framework model space. By solving the Kriging equation system, several equally probable random parameter fields corresponding to different key parameters are generated. The random parameter fields corresponding to different key unknown parameters constitute the heterogeneous stochastic geological model. In this embodiment, 50 heterogeneous stochastic geological models are generated. The sequential Gaussian collaborative simulation method with soft data constraints generates multiple heterogeneous stochastic geological models. This method can fully consider the randomness of geological modeling in three-dimensional space, and the generated stochastic models have strong heterogeneity, clear geological significance, and are more realistic.

[0024] S3: For each candidate site, select multiple representative models from its heterogeneous random geological models, perform multiphysics coupled numerical simulations on the representative models, calculate the performance indices of different representative models, and form a performance index set for each identical performance index of different representative models. Furthermore, to balance computational cost and statistical representativeness, 5-10 of the most representative models are selected from the generated heterogeneous stochastic geomechanical field models for subsequent simulations. The specific steps of the multiphysics coupled numerical simulation are as follows: S31: The geological architecture model is divided into grids, and the random parameter field of each representative model is mapped to each grid cell of the geological architecture model to obtain the corresponding simulation parameter field model. Furthermore, the random parameter fields in the heterogeneous random geological model, including the permeability random field K(x, y, z), the elastic modulus random field E(x, y, z), the cohesion random field c(x, y, z), and the internal friction angle random field β(x, y, z), are mapped to each grid cell through interpolation. For example, the griddata function in the scipy.interpolate module of the Numpy extension library can be used for interpolation.

[0025] S32: Establish the solid deformation relationship, gas permeation relationship, and component diffusion relationship for each representative model to obtain the corresponding set of governing equations; Furthermore, the specific method for establishing the solid deformation relationship is as follows: An incremental small deformation assumption method is used to establish the equilibrium equation, which is: In the formula, Where p is the effective stress tensor, α is the pore pressure, α is the Biot coefficient obtained from core laboratory test data, with a value ranging from 0.5 to 1, and I is the unit tensor. The bulk density is obtained from laboratory core testing data, and g is the gravitational acceleration vector; the displacement-strain geometric equation is established: In the formula, ε is the strain tensor and u is the displacement vector; based on the elastic modulus random field mapped to the geological framework model, and based on the elastoplastic constitutive relation, the stress-strain relationship between the strain tensor in the displacement-strain geometric equation and the effective stress tensor in the equilibrium equation is established, thereby obtaining the solid deformation relationship between the associated displacement and the effective stress. Preferably, the elastoplastic constitutive model satisfies the Drucker-Prager criterion, and its yield function F is: Where q is the deviatoric stress, and the plastic strain increment in the elastoplastic constitutive model is determined by the associated plastic flow rule; The specific method for establishing the gas permeation relationship is as follows: Establish the permeation-compressibility correlation equation for hydrogen gas in porous media: In the formula, φ represents porosity. The hydrogen density is obtained through the functional relationship between hydrogen density and compressibility factor, pressure, and temperature in the real gas effect of hydrogen. For example, the hydrogen density can be obtained by calling the CoolProp library, where t is the time variable, v is the Darcy velocity vector, and it satisfies the Darcy law formula. μ is the dynamic viscosity of hydrogen gas, and K is the permeability. In this embodiment, the dynamic viscosity of hydrogen gas is considered a constant. To represent the source and sink terms of the injection-production well; substituting Darcy's law into the seepage-compressibility correlation equation, we obtain a nonlinear gas seepage relationship with pore pressure p as the variable: Among them, porosity Changes and volumetric strain It is related to the pore pressure p, that is, Volumetric strain It is derived from the strain tensor ε. It is the bulk modulus of solid particles, derived from the random field of elastic modulus; The component diffusion relationship reflects the dissolution and diffusion law of hydrogen in pore water. The specific method for establishing the component diffusion relationship is as follows: Based on Fick's law governing the molecular diffusion of hydrogen in residual formation water, the component diffusion relationship is established as follows: ; In the formula, Let C be the water saturation, whose value is related to pore pressure. The relationship between water saturation and pore pressure satisfies the existing water retention curve. Let C be the hydrogen concentration in the water, and D be the effective diffusion coefficient tensor, the value of which is determined by water saturation and porosity. R is the aqueous phase seepage velocity, which is usually very small and is ignored in this embodiment. R is the chemical reaction source term, which is assumed to be 0 in this embodiment as there is no chemical reaction.

[0026] S33: Based on the simulation parameter field model, the sequential coupling iterative method is used to solve the corresponding control equations of each representative model within the complete simulation cycle to obtain the corresponding coupled simulation results. Furthermore, the sequential coupling iterative method can employ an implicit iterative method, and the time discretization uses the first-order implicit Euler method to ensure stability, specifically as follows: Set boundary conditions and operating conditions, including the initial stress field, the initial pore pressure field, and the applied cyclic injection and production pressure loads; A complete simulation cycle is divided into several time steps, and in each time step... The simulation is performed multiple times, n=1,2,3,…,Mt, where Mt is the total number of time steps. A complete simulation cycle is, for example, 10 injection-mapping cycles. Each iteration fixes the displacement field obtained from the previous time step. Or the displacement field obtained from the previous iteration at the current time step. k = 1, 2, 3, ..., Mk, where Mk is the upper limit of the iteration number. The problem is solved by coupling the gas permeation relationship and the component diffusion relationship. The pore pressure field and hydrogen concentration field at the current time step are updated iteratively to obtain the pore pressure field at the current time step and the current iteration number. and hydrogen concentration field Since the coupling relationship between gas permeation and component diffusion is nonlinear, this embodiment uses the Newton-Raphson method for iterative solution. Utilizing the updated pore pressure field and hydrogen concentration field Solve for the solid deformation relationship, update the displacement field at the current time step, and obtain the displacement field at the current time step and the current iteration number. ; After each iteration, a convergence check is performed. The convergence is determined by whether the increment norms of the pressure field, displacement field, and hydrogen concentration field are less than the set tolerance. If convergence is not achieved, the next iteration of the current time step is performed. If convergence is achieved, the next time step is entered. The coupled simulation results obtained after iteratively solving the control equations include the pore pressure field, displacement field, and hydrogen concentration field that vary with time step. Furthermore, during the simulation process, the output curves of wellhead pressure-time and hydrogen flow rate-time are generated.

[0027] S34: Perform post-processing calculations on the corresponding coupled simulation results of each representative model to obtain different performance indices. Each identical performance index of different representative models forms a performance index set.

[0028] Furthermore, the series of performance indicators includes sealing performance indicators. Injectability indicators Stability indicators Hydrogen purity maintenance index The sealing performance index is measured at the outer boundary of the cap layer. The result is obtained by integrating the change in hydrogen leakage flux over time, that is,

[0029] In the formula, For a complete simulation cycle, This refers to the total mass of hydrogen injected. The velocity vector field is obtained from the pore pressure field based on the simulation solution, and then calculated using Darcy's law. The unit outward normal vector of the outer boundary of the cap layer; The injectability metric is obtained by directly obtaining the simulation time required to reach the target capacity or by calculating the average injection impedance. We obtained, among which, The average pressure difference between the wellhead and the reservoir is extracted from the digital logging sequence. The average hydrogen injection rate is extracted from the wellhead pressure-time and hydrogen flow rate-time curves output from the simulation solution process. The stability index includes a damage index and a volume stability index. The damage index is the maximum equivalent plastic strain of the surrounding rock of the reservoir (e.g., the unit near the inner wall of the reservoir) during the entire simulation period. The maximum plastic strain is obtained based on the displacement field obtained from the simulation and the stress-strain relationship. The volume stability index is calculated as follows: the displacement of the cavity boundary nodes is calculated based on the displacement field and the volume is reconstructed to obtain the cavity volume. The volume change rate of the reservoir cavity is calculated to characterize the volume stability index. The method for calculating the hydrogen purity maintenance index is as follows: based on the pore pressure field, displacement field, and hydrogen concentration field obtained from simulation, the component flow rate at the extraction end is tracked during the simulation process, and the mole fraction of hydrogen in the extracted gas at the end of the extraction stage is calculated. , and the purity of hydrogen injection In comparison, the hydrogen purity maintenance index was calculated. .

[0030] Furthermore, the same performance metric corresponding to each representative model forms a performance metric set. For example, the sealing performance metrics of different representative models can form a sealing performance metric set. N represents the number of representative models for each candidate library address.

[0031] Furthermore, the multiphysics coupling simulation process adopts a stochastic simulation method to simulate the real cyclic injection and production process of a real compressed hydrogen energy storage underground reservoir in different heterogeneous stochastic geological models, thereby obtaining a set of performance indicators with non-unique values. The simulation results are more comprehensive and realistic, and the performance indicators are all physical quantities that can be directly used for engineering design, which is more instructive for site selection assessment.

[0032] S4: For each candidate repository, perform uncertainty quantification on its performance index set. Based on the uncertainty quantification result, use a multi-objective optimization ranking method to comprehensively score and rank each candidate repository to obtain the preferred repository with the highest ranking.

[0033] Furthermore, the specific method for uncertainty quantification is as follows: the mean, standard deviation and probability distribution of each performance index set are statistically analyzed to obtain the uncertainty quantification results of different performance indicators, thereby enabling the prediction of engineering risks brought about by geological uncertainty and directly transmitting geological uncertainty to engineering performance prediction. The multi-objective optimization ranking method specifically involves minimizing the sealing performance index, injectability index, stability index, and maximizing the hydrogen purity maintenance index as optimization objectives. It utilizes an approximation-ideal-solution ranking method or a random forest-based regression model, combined with the decision-maker's preferences, to comprehensively score multiple performance indicators for each candidate site. The preferred multi-objective optimization ranking method is the approximation-ideal-solution ranking method. A risk assessment is performed on the top-ranked preferred site, providing predictions of its average performance indicators and risk probability distribution maps for each performance indicator. This clearly indicates the performance boundaries and risk probabilities at a specific confidence level, thus providing a transparent and solid basis for the final decision. The site selection decision is more scientific, transparent, and traceable, significantly reducing human intervention and subjective bias.

[0034] Taking a planned salt cavern hydrogen storage project as an example, there are three exploratory wells in each of the two candidate sites, A and B. The site selection process for the underground storage of this salt cavern hydrogen storage project is as follows: Step 1: Collect high-resolution scan images of cores, digital logging sequences, and core laboratory test data from three exploration wells in candidate reservoir sites A and B. Use convolutional neural networks to train the high-resolution scan images of the cores from both candidate sites A and B, identifying and quantifying a hard gypsum interlayer in candidate site A, and then extracting its thickness and frequency characteristics as part of the micro-stratigraphic features of candidate site A. Analyze the digital logging sequences of both candidate sites A and B using long short-term memory networks, obtaining the following results: Figure 3 The macroscopic stratigraphic features of the two candidate reservoir sites, A and B, show that three regional mudstone strata exist in candidate site A and one regional mudstone strata exist in candidate site B. Principal component analysis was used to extract the parameter correlation features of candidate sites A and B. Among them, the principal components of sonic transit time logging and density logging in candidate site A are highly correlated with laboratory permeability.

[0035] Step 2: Based on the analysis results of the digital logging curve sequences of candidate reservoir sites A and B, geological framework models for candidate reservoir sites A and B are constructed respectively. Within the geological framework models of candidate reservoir sites A and B, the sequential Gaussian co-simulation method is used, with the frequency of anhydrite interlayers as soft data, to simulate 50 random fields of cohesion and internal friction angle of salt rock layers. For candidate reservoir site A, sonic transit time logging data and density logging data obtained from principal component analysis are used as soft data, and for candidate reservoir site B, resistivity and natural gamma logging data obtained from principal component analysis are used as soft data to simulate 50 random fields of permeability, thereby generating 50 corresponding heterogeneous stochastic geological models.

[0036] Step 3: From 50 heterogeneous stochastic geological models of candidate reservoir sites A and B, select 5 models as representative models for each. Perform multiphysics coupled numerical simulations on each representative model to obtain different performance indices. Candidate site A, due to its numerous interlayers, has lower caprock sealing and stability indices, but greater fluctuations in injectability. Conversely, candidate site A, with a more homogeneous caprock, has better injectability, but... Figure 4 The simulation results of one of the representative models of the candidate reservoir site region A shown indicate that local hydrogen accumulation and slight leakage occurred at the edge of the mudstone interlayer.

[0037] Step 4: Set the decision weights as follows: 40% for sealing performance, 30% for injectability, 20% for stability, and 10% for hydrogen purity maintenance. After comprehensive ranking using the approximation ideal solution ranking method, the comprehensive score table shown in Table 1 is obtained. The results show that candidate site A has a higher comprehensive score and is the preferred site. Table 1 Comprehensive Scoring Table

[0038] A risk assessment was conducted on candidate site A, resulting in the preferred site risk assessment table shown in Table 2. The results show that the sealing risk of candidate site A is extremely low, but there is a certain risk of fluctuation in injection and production efficiency. The quantitative comprehensive scoring and risk assessment results provide a clear basis for the site selection decision of this project.

[0039] Table 2 Risk Assessment Table for Preferred Sites

[0040] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for site selection of an underground compressed hydrogen energy storage facility, characterized in that, Includes the following steps: S1: Collect multi-source borehole data for each candidate reservoir site, including high-resolution core scanning images, digital logging curve sequences, and core laboratory test data, and extract the microscopic lithological characteristics, macroscopic stratigraphic characteristics, and parameter correlation characteristics of each candidate reservoir site from them. S2: For each candidate reservoir site, a corresponding geological framework model is constructed based on its macroscopic stratigraphic characteristics. Within the corresponding geological framework model, the sequential Gaussian co-simulation method is adopted, using its core laboratory test data as hard data and its data reflecting microscopic lithological characteristics and parameter correlation characteristics as soft data, to generate several heterogeneous stochastic geological models containing stochastic parameter fields. S3: For each candidate site, select multiple representative models from its heterogeneous random geological models, perform multiphysics coupled numerical simulations on the representative models, calculate the performance indices of different representative models, and form a performance index set for each identical performance index of different representative models. S4: For each candidate repository, perform uncertainty quantification on its performance index set. Based on the uncertainty quantification result, use a multi-objective optimization ranking method to comprehensively score and rank each candidate repository to obtain the preferred repository with the highest ranking.

2. The method for site selection of an underground compressed hydrogen energy storage facility according to claim 1, characterized in that, The specific steps of S3 are as follows: S31: The geological architecture model is divided into grids, and the random parameter field of each representative model is mapped to each grid cell of the geological architecture model to obtain the corresponding simulation parameter field model. S32: Establish the solid deformation relationship, gas permeation relationship, and component diffusion relationship for each representative model to obtain the corresponding set of governing equations; S33: Based on the simulation parameter field model, the sequential coupling iterative method is used to solve the corresponding control equations of each representative model within the complete simulation cycle to obtain the corresponding coupled simulation results. S34: Perform post-processing calculations on the corresponding coupled simulation results of each representative model to obtain different performance indices. Each identical performance index of different representative models forms a performance index set.

3. The method for site selection of an underground compressed hydrogen energy storage facility according to claim 2, characterized in that... The method for establishing the solid deformation relationship in S32 is as follows: The equilibrium equations are established using the incremental small-deformation assumption method, and the equilibrium equations are as follows: In the formula, The effective stress tensor is given by p, the pore pressure is given by α, the Biot coefficient is given by I, and the unit tensor is given by I. Let g be the volume density, and g be the gravitational acceleration vector. Establish the displacement-strain geometric equations: In the formula, ε is the strain tensor and u is the displacement vector; Based on the elastoplastic constitutive relation, the stress-strain relationship between the strain tensor and the effective stress tensor is established, and the solid deformation relationship between the associated displacement and the effective stress is obtained.

4. The method for site selection of an underground compressed hydrogen energy storage facility according to claim 3, characterized in that... The method for establishing the gas permeation relationship in S32 is as follows: Establish the percolation-compressibility correlation equation for hydrogen gas in porous media: In the formula, φ represents porosity. Let be the density of hydrogen gas, t be the time variable, and v be Darcy's velocity vector, satisfying Darcy's law formula. μ is the dynamic viscosity of hydrogen gas, and K is the permeability. This represents the source and sink terms of the injection and production wells; Substituting Darcy's law into the seepage-compressibility correlation equation, we obtain a nonlinear gas seepage relationship with pore pressure p as the variable: Among them, the change in porosity φ is related to volumetric strain. It is related to the pore pressure p, that is, , It is the bulk modulus of solid particles.

5. The method for site selection of an underground compressed hydrogen energy storage facility according to claim 4, characterized in that, The method for determining the component diffusion relationship in S32 is as follows: Based on Fick's law governing the molecular diffusion of hydrogen in residual formation water, the component diffusion relationship is established: ; In the formula, Let C be the water saturation, C be the hydrogen concentration in the water, and D be the effective diffusion coefficient tensor. R represents the aqueous phase seepage velocity, and R is the chemical reaction source term, with a value of 0.

6. The method for site selection of an underground compressed hydrogen energy storage facility according to claim 5, characterized in that, The sequential coupling iterative method in S33 is specifically as follows: Set boundary conditions and operating conditions; A complete simulation cycle is divided into several time steps, and multiple iterations are performed in each time step. Each iteration fixes the displacement field obtained in the previous iteration at the current time step, and solves the problem by coupling the gas permeation relationship and the component diffusion relationship, and iteratively updates the pore pressure field and hydrogen concentration field at the current time step. Using the updated pore pressure field and hydrogen concentration field, the solid deformation relationship is solved, and the displacement field at the current time step is updated. Iterate multiple times at the current time step until convergence, then proceed to the next time step; The coupled simulation results obtained after iteratively solving the output control equations include the pore pressure field, displacement field, and hydrogen concentration field that vary with the time step.

7. The method for site selection of an underground compressed hydrogen energy storage facility according to claim 6, characterized in that, The specific method for post-processing calculation in S34 is as follows: The velocity vector field is calculated based on the pore pressure field, and the sealing index is calculated by combining the total hydrogen injection mass. During the simulation solution process, the wellhead pressure-time curve and hydrogen flow-time curve are output. The average hydrogen injection rate is extracted from them, and the average pressure difference between the wellhead and the reservoir is extracted from the digital logging curve sequence. The injectability index is then calculated. The maximum equivalent plastic strain and the rate of change of cavity volume are calculated based on the displacement field throughout the entire simulation period. Based on the pore pressure field, displacement field, and hydrogen concentration field, the component flow rate at the extraction end is tracked during the simulation process, and the hydrogen purity maintenance index is calculated.

8. The method for site selection of an underground compressed hydrogen energy storage facility according to claim 7, characterized in that... The specific method for uncertainty quantification in S4 is as follows: the mean, standard deviation and probability distribution of each performance index set are statistically analyzed to obtain the uncertainty quantification results of different performance indices.

9. The method for site selection of an underground compressed hydrogen energy storage facility according to claim 8, characterized in that... The multi-objective optimization ranking method in S4 is as follows: with the optimization objectives of minimizing the sealing performance index, minimizing the injectability index, minimizing the stability index, and maximizing the hydrogen purity maintenance index, the ranking method of approximating the ideal solution is used, combined with the decision-maker's preferences, to comprehensively score multiple performance indicators of each candidate site. A risk assessment is conducted on the top-ranked preferred site, its average performance indicators are predicted, and a risk probability distribution map of each performance indicator is provided.

10. The method for site selection of an underground compressed hydrogen energy storage facility according to claim 1, characterized in that... In S1, The specific method for extracting microscopic lithological features is as follows: a convolutional neural network is used to train high-resolution scanned images of rock cores to automatically identify and extract microscopic lithological features; The specific method for extracting macroscopic stratigraphic features is as follows: Long Short-Term Memory (LSTM) network is used to analyze continuous digital logging curve sequences to automatically identify macroscopic stratigraphic features; The specific method for extracting parameter correlation features is as follows: principal component analysis is performed on the core laboratory test data and the corresponding logging response data in the digital logging curve sequence to extract the comprehensive characteristic factors that can best represent the changes in rock physical and mechanical properties as parameter correlation features.

Images